A.D. Varlamov - Ph.D. (Eng.), Vladimir State University named after Alexander and Nikolay Stoletovs. E-mail: varlamov_aleks@mail.ru R.V. Sharapov - Ph.D. (Eng.), Vladimir State University named after Alexander and Nikolay Stoletovs. E-mail: info@vanta.ru

There are a number of specific task in area monitoring. One of them is to estimate the area covered with water. This problem is important in assessing the river floods, creating congestion and dams, fast flowing rivers assessment, identify newly formed or dried bodies of water (waterlogging level), etc. As background information for the assessment of large areas in length can be used remote sensing (satellite observations, aerial photography, etc.). In connection with great extent of the territory of our country and the big volume of satellite data, the actual tasks are to conduct automated analysis of satellite imagery and implementation of site assessment without human intervention. The problem can be solved by the use of digital image processing and image analysis. The problem is formulated as follows: for each point color satellite image must determine whether the corresponding portion of the area to the water surface at the time the photographs. Since the task set is described in a general way and cannot be formalized, segmentation method will be based on machine learning. To realize the image segmentation algorithm aqueous or non-aqueous portions following steps are performed: 1. formed collection of satellite images, which may be present bodies of water; 2. the human expert allocated water areas. In conjunction with the initial training sample collection is obtained; 3. implements the image processing and analysis to calculate the values of local and global features; 4. running machine learning (learning classifier); 5. learning outcomes in the form of an algorithm in conjunction with an algorithm for computing features introduced in the system. In this paper was used as the classifier formal neural network.The proposed solutions have been successfully used to estimate the Basin of Oka River in Nizhny Novgorod and Vladimir regions.

 

1. Varlamov A.D. Vosstanovlenie cveta polutonovykh izobrazhenijj nejjronnojj setju // Algoritmy, metody i sistemy obrabotki dannykh. 2011. Vyp. 2(17). S. 55-62.
2. Varlamov A.D., SHarapov R.V. Ispolzovanie nejjronnykh setejj v zadachakh monitoringa ehkzogennykh processov distancionnymi metodami // Geoinformatika. 2014. № 4. C. 62-68.
3. Kokoulin A.N. Razrabotka ehffektivnykh metodov nadezhnogo khranenija i peredachi mnogomernykh massivov graficheskikh dannykh v informacionno-upravljajushhikh sistemakh // Nejjrokompjutery: razrabotka, primenenie. 2013. № 11. S. 70-75.
4. Koljuchkin V.JA., Nguen K.M., CHan T.KH. Algoritmy obrabotki izobrazhenijj v sistemakh mashinnogo zrenija robotizirovannykh proizvodstvennykh linijj //Nejjrokompjutery: razrabotka, primenenie. 2014. № 3. S. 44-51.
5. Noskov A.A., Aminova E.A., Priorov A.L., Trapeznikov I.N. Sistema opredelenija povtornogo detektirovanija obektov na osnove nejjronnojj seti // Nejjrokompjutery: razrabotka, primenenie. 2014. № 3. S. 36-43.
6. Romanova A.G., Tarkhov D.A., SHemjakina T.A. Nejjrosetevoe modelirovanie ehkologii // Nejjrokompjutery: razrabotka, primenenie. 2014. № 2. S. 16-21.
7. SHarapov R.V., Varlamov A.D.Sravnitelnyjj analiz sistem poiska graficheskikh dannykh // Sovremennye naukoemkie tekhnologii. 2013. № 1. S. 27-31.
8. SHarapov R.V. Organizacija avtomaticheskogo nabljudenija za sostojaniem poverkhnostnykh vod // Mashinostroenie i bezopasnost zhiznedejatelnosti. 2014. №2(20). S. 32-38.
9. Sharapov R., Varlamov A. Machine learning of visually similar images search // CEUR Workshop Proc. 2012. V. 934.  P. 113-120.